Hyper-V and DFS...what if...

Witor · 16 Jul 2009 at 11:50

Hi all,

Ok a scenario, currently running 2 servers, Svr1 and Svr2 as fileservers, in a DFS system
(Windows 2008, DFS, 2x quad core Xeons, 16GB ram each, both servers have 4x450 SCSI drives in a RAID 5 config, internal bays (IBM x3650's), OS on each server resides on 2x300GB SCSI, Raid1)

Obviously DFS replicates the documents onto both servers as usual, this is working perfectly, and has been for a while now.

Now..both servers are capable of running Hyper-V, which leads me to the following:

Say I install Hyper-V on both Svr1 and Svr2..

If I run a Virtual server on Svr1 via Hyper-V, how does DFS manage the replication of the VHD image? (Got 4 Virtual servers who's images range from 2.5GB to 25GB...) is it at byte level? does it copy the whole vhd every time a change is made?

Im looking at basically having a High Availability solution, in the event that Svr1 goes **** up, the virtual servers VHD files are already on Svr2, and can be loaded up relatively quickly...

thoughts?

Rationale behind this is that we have a server dedicated for Hyper-V (call it Svr3) at the moment, running 4 virtual servers (a 2003 print server, a sophos antivirus enterprise server, a websense reporting server and a fedora server running naigos). If I could move these virtual servers onto Svr1, it would free up Svr3 for other uses....

Saundie · 16 Jul 2009 at 13:14

Does DFS replicate changes on open files? From what I recall it is not capable of doing so, which means that your strategy will not work as the VHDs will be in constant use. I am finding it difficult to confirm this though - I will update this post once I am sure.

Witor · 16 Jul 2009 at 13:19

Think youre right off memory...will look into that too..

Rangler · 16 Jul 2009 at 14:20

From memory, and its been a while but im sure 2008 uses RDC as standard
(Remote Differential Compression) so it would only replicate changed data.

However also im sure it works on closed files only, replicating changes in open files could lead to inevitable corruption.
However if you were to take snapshots regularly I suppose you could replicate the snapshots and you would be very close to what your trying to achieve without going down the ESX HA route.

Saundie · 16 Jul 2009 at 14:46

Personally I'd go for a proper cluster using iSCSI to get HA for Hyper-V. If you go that route, you get proper tools to manage the VMs, and Live Migration if you upgrade to R2 when it's released in a few months time. iSCSI target isn't too expensive - I used StarWind in my last implimentation, which is about £400 per server these days. At least this way, you get a clean, robust solution.

Witor · 16 Jul 2009 at 15:19

Yep, planning to upgrade to R2 when it's released.

Problem is, there are no funds as such at the moment for stuff like Starwind, which I have looked at (and seemed to fit the bill)

Thanks for the answers, it has cleared a few things up.

Saundie · 16 Jul 2009 at 16:06

Yes, that's something I encountered when planning my Hyper-V deployment, in the end we used disaster recovery as a way to secure additional budget.

I do not know what your setup is like, but we tried using OpenFiler which is an open source SAN/NAS appliance on an old server as a proof of concept, so if you have a spare box lying around you could give it a try and see if it suits your needs.

Anyway, best of luck with it